Title: Comparison of MRI scanning protocol in terms of quantitative MRI measures

Question 1: Which protocols had large numbers of subjects?

Protocol / Coil name / Pixel spacing / Repetition time / Pixel band width / Number of scans
1 / Head / 0.9375 / 3000 / 84.9219 / 2072
2 / Head / 0.9375 / 3000 / 81.40625 / 32
3 / Head / 0.9375 / Not 3000 / 84.9219 / 66
4 / 8 HR Brain / 0.9375 / 3000 / 84.9219 / 52
5 / 8 HR Brain / 0.8594 / 3000 / 84.9219 / 85
6 / 8 HR Brain / 0.8594 / 3000 / 81.4062 / 47
7 / 8 HR Brain / 0.8594 / Not 3000 / 84.9219 / 158
8 / 8 HR Brain / 0.8594 / Not 3000 / 81.4062 / 54
9 / 8 HR Brain / 0.8594 / Not 3000 / 244.141 / 25
10 / 8 HR Brain / 0.7813 / 3000 / 84.9219 / 185
11 / 8 HR Brain / 0.7813 / 3000 / 81.4062 / 87
12 / 8 HR Brain / 0.7813/0.7812 / Not 3000 / 84.9219 / 119
13 / 8 HR Brain / 0.7813/0.7812 / Not 3000 / 81.4062 / 23
14 / 8 HR Brain / 0.7813/0.7812 / Not 3000 / 244.141 / 25

The classification not 3000 for the variable named REPETITIONTIME1 was used since very few other repetition times had a large number of subjects.

CROSS-SECTIONAL ANALYSIS

Question 2: What is the mean (SD) BPF, T2 lesion volume and ICC volume at first visit?

Protocol / BPF_TDS / BPF_MAN / T2LV_TDS / T2LV_MAN / ICC_TDS / ICC_MAN
1 / 0.86+/-0.046 / 0.849+/-0.048 / 4.214+/-3.579 / 6.082+/-6.87 / 1391.796+/-330.477 / 1375.982+/-142.497
2 / 0.856+/-0.042 / 0.842+/-0.044 / 6.79+/-6.502 / 10.634+/-11.121 / 1375.285+/-137.779 / 1376.396+/-138.851
3 / 0.864+/-0.046 / 0.85+/-0.053 / 6.149+/-7.876 / 10.516+/-15.99 / 1387.831+/-144.1 / 1387.313+/-143.746
4 / 0.864+/-0.049 / 0.86+/-0.044 / 3.417+/-3.241 / 5.18+/-5.845 / 1382.64+/-133.704 / 1382.007+/-134.441
5 / 0.838+/-0.048 / 0.822+/-0.05 / 6.01+/-5.061 / 9.87+/-11.608 / 1572.89+/-231.497 / 1435.752+/-146.78
6 / 0.846+/-0.06 / 0.819+/-0.08 / 9.206+/-9.332 / 14.781+/-19.743 / 1756.756+/-242.314 / 1506.122+/-187.679
7 / 0.88+/-0.052 / 0.848+/-0.055 / 6.56+/-3.914 / 8.091+/-10.289 / 1594.845+/-184.14 / 1405.107+/-132.007
8 / 0.858+/-0.057 / 0.839+/-0.059 / 5.079+/-3.992 / 6.126+/-6.977 / 1670.119+/-191.768 / 1414.447+/-132.185
9 / 0.866+/-0.044 / 0.853+/-0.049 / 2.758+/-2.467 / 3.477+/-3.811 / 1642.705+/-174.081 / 1381.483+/-148.179
10 / 0.857+/-0.052 / 0.833+/-0.057 / 7.404+/-5.24 / 8.236+/-8.945 / 1781.241+/-369.254 / 1385.1+/-140.579
11 / 0.858+/-0.044 / 0.839+/-0.051 / 7.146+/-4.708 / 6.638+/-6.258 / 1853.241+/-234.832 / 1352.75+/-111.694
12 / 0.896+/-0.045 / 0.855+/-0.047 / 7.857+/-4.449 / 6.645+/-8.055 / 1681.493+/-318.776 / 1347.173+/-113.933
13 / 0.883+/-0.03 / 0.861+/-0.031 / 6.473+/-2.994 / 5.577+/-4.748 / 1861.523+/-252.486 / 1363.914+/-83.432
14 / 0.837+/-0.056 / 0.824+/-0.06 / 4.365+/-3.587 / 5.147+/-5.354 / 1765.412+/-323.868 / 1334.979+/-112.273

Question 3: What is the difference between each protocol and protocol 1?

Protocol / BPF_TDS / BPF_MAN / T2LV_TDS / T2LV_MAN / ICC_TDS / ICC_MAN
2 v 1 / -0.00407 / -0.00655 / 2.575773 / 4.552112 / -16.5109 / 0.413637
3 v 1 / 0.004166 / 0.00096 / 1.935555 / 4.434291 / -3.96511 / 11.33102
4 v 1 / 0.004294 / 0.011692 / -0.7969 / -0.90203 / -9.15548 / 6.024821
5 v 1 / -0.02174 / -0.0265 / 1.79646 / 3.788726 / 181.0938 / 59.7703
6 v 1 / -0.01365 / -0.02937 / 4.992613 / 8.698996 / 364.9608 / 130.1404
7 v 1 / 0.020317 / -0.00038 / 2.346055 / 2.008934 / 203.0489 / 29.12466
8 v 1 / -0.00253 / -0.00987 / 0.865234 / 0.044166 / 278.3237 / 38.46505
9 v 1 / 0.006322 / 0.004747 / -1.4554 / -2.60467 / 250.9097 / 5.501159
10 v 1 / -0.00332 / -0.01542 / 3.189931 / 2.15462 / 389.4453 / 9.117627
11 v 1 / -0.00194 / -0.00928 / 2.931698 / 0.556556 / 461.445 / -23.2319
12 v 1 / 0.035433 / 0.006027 / 3.643176 / 0.563683 / 289.6977 / -28.8088
13 v 1 / 0.022825 / 0.012101 / 2.258643 / -0.50493 / 469.7268 / -12.0678
14 v 1 / -0.0233 / -0.02504 / 0.15081 / -0.93493 / 373.6163 / -41.0033

Question 4: Was there a significant difference in the variance across the protocols?

BPF_TDS: p=0.265

BPF_MAN: p=0.011

log(T2LESION_TDS): p=0.021

log(T2LESION_MAN): p=0.125

ICC_TDS: p<0.000001

ICC_MAN: p=0.126

Question 5: What are the estimated differences among the MS subtypes using just protocol 1 or using all protocols with an additive offset?

Group comparison / BPF_TDS/ protocol 1 / BPF_TDS/ all other protocols / BPF_MAN/ protocol 1 / BPF_MAN/ all other protocols
SP v RR / -0.02711 / -0.02929 / -0.03261 / -0.03829
PP v RR / 0.00085 / -0.02132 / 0.001775 / -0.02078
PR v RR / -0.04505 / -0.05969 / -0.03979 / -0.05777
CIS v RR / 0.016525 / 0.014257 / 0.018262 / 0.018555
Group comparison / T2LV_TDS/ protocol 1 / T2LV_TDS/ all protocols / T2LV_MAN/ protocol 1 / T2LV_MAN/ all protocols
SP v RR / 1.739442 / 2.280568 / 4.303779 / 7.105516
PP v RR / -1.41287 / -0.16625 / -1.84471 / -0.36118
PR v RR / 2.76516 / 0.241796 / 7.431093 / 3.209743
CIS v RR / -1.09853 / -1.95398 / -2.97371 / -2.6338
Group comparison / log(T2LV_TDS)/ protocol 1 / log(T2LV_TDS)/ all protocols / log(T2LV_MAN)/ protocol 1 / log(T2LV_MAN)/ all protocols
SP v RR / 0.335122 / 0.308298 / 0.574774 / 0.691455
PP v RR / -0.37474 / 0.133734 / -0.45067 / 0.122516
PR v RR / 0.748117 / 0.101198 / 1.245117 / 0.632075
CIS v RR / -0.30379 / -0.38032 / -0.66203 / -0.5503
Group comparison / ICC_TDS/ protocol 1 / ICC_TDS/ all protocols / ICC_MAN/ protocol 1 / ICC_MAN/ all protocols
SP v RR / 5.789286 / 22.00275 / -1.60813 / 14.41535
PP v RR / -16.744 / 50.49123 / -39.8898 / 24.35706
PR v RR / -42.4733 / -166.063 / -1.28069 / 113.1067
CIS v RR / -9.28709 / -53.7328 / 14.90533 / 11.6525

Question 6: Was there a significant protocol by group interaction in any of the cross-sectional analyses?

The goal of this analysis was to assess whether the differences between groups might be different across the protocols. Limited information was available for this analysis because not all groups were observed on all of the protocols.

BPF_TDS: p=0.145

BPF_MAN: p=0.033

T2LESION_TDS: p=0.003

T2LESION_MAN: p=0.0006

log(T2LESION_TDS): p=0.263

log(T2LESION_MAN): p=0.313

ICC_TDS: p=0.745

ICC_MAN: p=0.149

Question 7: What are the estimated differences between males and females using just protocol 1 or using all protocols with an additive offset?

Male v female/ TDS protocol 1 / Male v female/ TDS other protocols / Male v female/ MAN protocol 1 / Male v female/ MAN other protocols
BPF / -0.02363 / -0.0244 / -0.02282 / -0.02185
T2LV / 1.236011 / -0.24185 / 1.950238 / -0.50508
log(T2LV) / 0.246294 / 0.06393 / 0.249596 / 0.071509
ICC / 165.8201 / 214.907 / 189.4137 / 164.9973

Question 8: Was there a significant protocol by gender interaction in any of the cross-sectional analyses?

The goal of this analysis was to assess whether the differences between males and females might be different across the protocols.

BPF_TDS: p=0.9313

BPF_MAN: p=0.987

T2LESION_TDS: p=0.105

T2LESION_MAN: p=0.614

log(T2LESION_TDS): p=0.832

log(T2LESION_MAN): p=0.878

ICC_TDS: p=0.361

ICC_MAN: p=0.159

Question 9: Was there a significant protocol by age interaction in any of the cross-sectional analyses?

BPF_TDS: p<0.0001

BPF_MAN: p<0.0001

T2LESION_TDS: p=0.002

T2LESION_MAN: p=0.002

ICC_TDS: p=0.281

ICC_MAN: p=0.316

Question 10: Which features of the protocol were significantly associated with each of the outcome measures?

BPF_TDS: PIXELBANDWIDTH=244.141 had significantly lower BPF than PIXELBANDWIDTH=84.9219

REPETITION TIME=3000 had significantly lower BPF than REPETITION TIME not equal to 3000

BPF_MAN: PIXELBANDWIDTH=244.141 had significantly lower BPF than PIXELBANDWIDTH=84.9219

REPETITION TIME=3000 had significantly lower BPF than REPETITION TIME not equal to 3000

Differences were smaller than for TDS

log(T2LESION_TDS): PIXELBANDWIDTH=244.141 had significantly lower T2LV than PIXELBANDWIDTH=84.9219 or PIXELBANDWIDTH=81.40625 and PIXELBANDWIDTH=84.9219 had significantly lower T2LV than PIXELBANDWIDTH=81.40625

PIXELSPACING=0.9375 had significantly lower T2LV than either of the other two PIXELSPACING values

log(T2LESION_MAN): PIXELBANDWIDTH=244.141 had significantly lower T2LV than PIXELBANDWIDTH=84.9219 or PIXELBANDWIDTH=81.40625

and PIXELBANDWIDTH=84.9219 had significantly lower T2LV than PIXELBANDWIDTH=81.40625

ICC_TDS: PIXELSPACING=0.9375 had significantly lower ICC than either of the other two PIXELSPACING values

ICC_MAN: No significant effects

LONGITUDINAL ANALYSES

Question 1: Is there a significant interaction between protocol and on study change in each of the measures using a random intercepts and slopes model?

BPF_TDS: p=0.0413 (PROTOCOL 12 had the largest difference)

BPF_MAN: p=0.0004 (PROTOCOL 4 and 6 had the largest differences).

T2LV_TDS: p=0.168

T2LV_MAN: p=0.063

ICC_TDS: p=0.402

ICC_MAN: p=0.235

Question 2: Is there a significant effect of protocol on the effect of disease category on the change with time?

BPF_TDS: p=0.069

BPF_MAN: p=0.815

T2LV_TDS: p=0.094

T2LV_MAN: p=0.284

ICC_TDS:

ICC_MAN: p=0.774

1